Phosphorylcholine, a crucial component of the pneumococcal cell wall, is essential in bacterial physiology and in human pathogenesis because it binds to serum components of the immune system and acts as a docking station for the family of surface choline‐binding proteins. The three‐dimensional structure of choline‐binding protein F (CbpF), one of the most abundant proteins in the pneumococcal cell wall, has been solved in complex with choline. CbpF shows a new modular structure composed both of consensus and non‐consensus choline‐binding repeats, distributed along its length, which markedly alter its shape, charge distribution and binding ability, and organizing the protein into two well‐defined modules. The carboxy‐terminal module is involved in cell wall binding and the amino‐terminal module is crucial for inhibition of the autolytic LytC muramidase, providing a regulatory function for pneumococcal autolysis.
The Gram‐positive bacterium Streptococcus pneumoniae is one of the most important human pathogens, which is commensal to the nasopharyngeal cavity. However, under appropriate conditions, pneumococci are able to cause invasive infections such as sepsis and meningitis (Cartwright, 2002). S. pneumoniae is unique among prokaryotes due to an absolute requirement for choline in growth, which is incorporated as phosphorylcholine into the cell wall teichoic acid and the membrane lipoteichoic acid (Tomasz, 1967). Choline is recognized by components of the host innate immune system, such as C‐reactive protein and the receptor of the platelet‐activating factor (Cundell et al, 1995). In addition, choline anchors a special class of pneumococcal proteins, referred to as choline‐binding proteins (CBPs), to the cell wall by non‐covalent interactions. These proteins are involved in a wide range of physiological functions, particularly adhesion to human cells (PspA, CbpA and CbpG; Bergmann & Hammerschmidt, 2006), as cell wall lytic enzymes (LytA, LytB, LytC and Pce; López & García, 2004), or in competence for genetic transformation (LytA, LytC and CbpD; Claverys & Havarstein, 2007). Many of these surface proteins are virulence factors that contribute to the pathogenesis of this organism. CBPs share a modular organization consisting of a biologically active domain and a conserved choline‐binding module (CBM). So far, full‐length three‐dimensional structures of CBPs are available for the phage‐encoded lysozyme Cpl‐1 (Hermoso et al, 2003) and pneumococcal Pce (Hermoso et al, 2005).
Here, we describe, for the first time, the crystal structure of CbpF, a member of a new subfamily of the CBPs that are strictly composed of a combination of canonical and non‐conserved choline‐binding motifs. CbpF is divided into two differentiated modules with specific mechanistic roles. Biochemical studies show that CbpF modulates the enzymatic activity of pneumococcal murein hydrolases, suggesting that this CBP could be essential in the regulation of these crucial enzymes.
Results And Discussion
Overall structure of the CbpF–choline complex
CbpF was crystallized, and its structure was solved by the single‐wavelength anomalous diffraction (SAD) technique. CbpF is folded into two well‐defined modules (Fig 1). The amino‐terminal module (residues 1–128) shows a disc‐shape conformation (37 Å × 34 Å × 24 Å). The CBM (residues 129–311) follows the superhelical fold previously found in other CBPs and is arranged into two regions referred to as CI and CII (Fig 1A). The CII domain (residues 193–311) presents five choline‐binding repeats (p1–p5) of around 20 amino acids each, with the consensus motif (GWXK‐X4−5‐WYY‐hydrophobic‐X3−5‐GXMX2−3), and a carboxy‐terminal tail of 16 amino acids (295–311). Each repeat comprises a symmetrical β‐hairpin, followed by a loop and a coiled region. The CI structural domain of CbpF (residues 129–192) presents two modified choline‐binding repeats (dp5 and dp6) showing additional amino acids inside the β‐hairpin of the dp6 repeat and also in the loop after the β‐hairpin of the dp5 repeat (coloured in red and cyan in Fig 1B). The CI domain defines a linker region between the N‐terminal module and the CBM.
The structure of CbpF has unique features compared with other CBPs, as it is assembled entirely by choline‐binding repeats. In the CII domain, the repeats strictly follow the consensus motif, whereas in the CI domain they are moderately modified by amino‐acid insertions. Notably, the repeats building the N‐terminal module are highly modified by both additional amino acids and mutations on different positions of the consensus motif (see Fig 1B). The presence of this choline‐binding repeat framework along the sequence produces the elongated shape of CbpF (>117 Å long) and results in a good structural superimposition of both canonical and modified repeats (Fig 1C).
The crystal structure of CbpF in complex with choline shows that up to seven choline‐binding sites are available (Fig 1). These seven choline molecules are all located at the CBM in such a way that at least three structurally conserved aromatic residues (two tryptophans from a repeat and one tyrosine from the next repeat) plus one hydrophobic residue (methionine or leucine) form a cavity in which choline methyl groups are placed (Fig 2A). This binding‐site framework is observed in both CI and CII domains; however, the N‐terminal module does not fulfil the choline‐binding requirements. We note that in dp1, dp2 and dp4 repeats at least one of the crucial tryptophan residues is replaced by threonine or lysine (Fig 1B). It would seem that this mutational replacement of residues produces a reduction in choline‐binding ability and that, in the case of the lysine residues, the formation of crucial salt bridge interactions affects modular rearrangement (see below).
Sequence divergence in the choline‐binding repeats
Sequence divergence observed in CbpF is generated by mutations of the consensus amino‐acid residues of the repeat, by amino‐acid additions in the β‐hairpin turn or the loop, or by a combination of both (Fig 1B). The high divergence of the first four repeats (dp1–dp4) is responsible for the absence of choline‐binding capacity at the N‐terminal region. Besides, modifications in the nature and composition of the choline‐binding repeats crucially alter other features of CbpF. Amino‐acid additions observed in dp3, dp4, dp5 and dp6 repeats (Fig 1B) result in longer repeats of 43, 39, 32 and 30 amino acids, respectively, compared with the 20 amino acids of a canonical choline‐binding repeat. Variations in the sequence of the choline‐binding repeats are also responsible for the modular arrangement in CbpF. Intermodular interactions are carried out through the N‐terminal module and the CI domain. Lys 90 is at a salt bridge interaction distance of Asp 89 (3.37 Å), and Lys 96 connects the N‐terminal module and the CBM (Fig 2B).
CbpF modulates the autolytic function of LytC
A highly sensitive and reliable method with radioactive pneumococcal cell wall as a substrate has been implemented to detect lytic activities of the pneumococcal cell wall. By using this assay, CbpF did not show any lytic activity, indicating that it is unlikely that CbpF has such enzymatic activity. Nevertheless, we tested whether CbpF could affect the activity of the known pneumococcal murein hydrolases (LytA, LytB, LytC and Pce). We assumed that CbpF could modulate the cell wall hydrolytic activities of these enzymes either by competing for the choline‐binding sites or for their substrate or by interacting with their catalytic or binding modules. To test this hypothesis, we determined the in vitro cell wall lytic activity of different murein hydrolases in the presence of CbpF or the isolated homologous CBM of LytA (C‐LytA). The results showed that LytA, LytB and Pce activities were not essentially affected by the addition of CbpF, but that the activity of the LytC lysozyme was reduced in the presence of 200 nM CbpF (supplementary Fig S2 online). To investigate the possible individual contribution of the isolated modules of CbpF to this inhibitory effect, N‐CbpF‐ (containing the first 192 amino acids) or C‐CbpF‐ (containing the CII domain; supplementary information online) truncated proteins were used in a similar radioactive in vitro experiment. C‐CbpF did not alter the enzymatic activity of LytC, suggesting that the N‐terminal domain of CbpF might be crucial in this specific inhibition. Nevertheless, the presence of the anchoring module of CbpF is required for its inhibitory function, because the isolated N‐CbpF was unable to inhibit the LytC activity (Fig 3A). These observations preclude the possibility that the inhibitory effect of CbpF on LytC could simply be ascribed to a competition for binding of choline residues of the cell surface, but the binding to choline residues is crucial to keep the CbpF protein in close contact with the cell wall. Although the N‐CbpF is able to bind to the peptidoglycan (see below), this interaction seems too weak to prevent the access of LytC to the substrate. This result agrees with the behaviour of other CBP murein hydrolases, in which the absence of the CBM reduces their functional activity to practically undetectable levels (Sanz et al, 1992).
To test whether the inhibitory effect of CbpF could affect other pneumococcal lysozymes and to confirm that this effect was not due to a binding competition for choline residues of cell wall, we assayed the pneumococcal phage lysozymes Cpl‐1 and Cpl‐7. Despite only the three‐dimensional structure of Cpl‐1 having been reported, sequence and biochemical analyses show that although the catalytic modules of both phage lysozymes are similar, the anchoring modules are completely different (Cpl‐7 does not bind to choline; García et al, 1990). Notably, both enzymes were inhibited by CbpF (Fig 3A), suggesting that the inhibitory effect is not limited to the LytC lysozyme and that the effect is not due to competition for choline residues.
As the LytC lysozyme is involved in the autolysis of pneumococcal cultures at 30°C (García et al, 1999), we used this property to investigate the role of CbpF in vivo. Therefore, we analysed the effect on autolysis of the absence or exogenous addition of CbpF in R6lytAcbpF, a double frameshift/insertion mutant of the wild‐type R6 S. pneumoniae strain. The mutant was LytA‐deficient to avoid the effect of this principal autolysin on the autolytic role of LytC at 30°C. The slow autolysis curve for R6lytA (López et al, 1986), a single frameshift mutant of R6, at 30°C was accelerated in the R6lytAcbpF strain (Fig 3B), suggesting that the absence of CbpF facilitates the autolytic process by LytC. This result agrees with the repression effect of LytC by CbpF observed in vitro, indicating that CbpF has the same regulatory function in vivo. Furthermore, when we analysed the growth of the R6lytA strain in the presence of CbpF added to the bacterial growth (i.e. the surface), we observed that the addition of 1.2 μM CbpF abolished LytC‐induced autolysis at 30 °C, whereas the addition of the truncated proteins C‐CbpF or C‐LytA did not alter the autolytic curve (Fig 3C). The culture containing exogenous CbpF showed the presence of typical unlysed R6lytA cells with a short‐chain cell morphology under phase‐contrast microscopy, whereas many lysed cells were observed in the culture without CbpF, indicating that CbpF protects the cell wall from lytic damage (data not shown).
Conversely, we tested the behaviour of R6lytAlytC, a double frameshift/insertion mutant, in the presence of LytC alone or LytC plus CbpF (Fig 3D). As expected, the mutant cells autolysed when they received a defined quantity of LytC (García et al, 1999); however, the autolysis was circumvented when LytC was added in the presence of 1.2 μM of CbpF. It is worth noting that the inhibition of this autolytic process did not take place in the presence of similar amounts of C‐CbpF or other homologous proteins such as LytB.
Taken together, all these results argue that CbpF selectively modulates the autolytic function of the LytC lysozyme and that the divergent choline‐binding repeats that build its N‐terminal domain have a crucial function in this effect. It is well established that autolysins have various fundamental roles in bacterial physiology and pathology (Gosink et al, 2000; López et al, 2004). These catalytically competent enzymes are able to hydrolytically change the integrity of the cell wall of the bacterium itself. The effect of uncontrolled reactions catalysed by these enzymes could be detrimental, and as such the enzymic activity must be tightly regulated. The nature of this regulatory process remains unknown, although there might be some contribution by the lipoteichoic acid in the cell membrane (Höltje & Tomasz, 1975). Recently, the involvement of small non‐coding RNAs in the control of stationary‐phase autolysis has also been suggested (Halfmann et al, 2007). Our results show that CbpF can inhibit the activity of LytC both in vitro and in vivo and, therefore, we assert that CbpF is crucial in the regulation of the activity of LytC.
The N‐terminal module and peptidoglycan interaction
Initial experiments to investigate a possible direct LytC–CbpF interaction, carried out by surface plasmon resonance and crosslinking tests, were negative. Therefore, the observed inhibition caused by CbpF on pneumococcal lysozymes could be due to specific interactions between CbpF and peptidoglycan, preventing access of the lysozyme to its substrate. To study this hypothesis, we analysed these interactions by using surface plasmon resonance—immobilizing cell walls from Micrococcus lysodeikticus and from S. pneumoniae with and without choline. As expected, CbpF was able to bind tightly to the pneumococcal choline‐containing cell wall. Interestingly, CbpF was also able to bind to the cell wall from S. pneumoniae without choline and, to a lesser extent, the cell wall of M. lysodeikticus, a standard Gram‐positive bacterial peptidoglycan that does not contain choline (Fig 4A). Furthermore, truncated C‐CbpF and C‐LytA proteins were not able to bind to the pneumococcal cell wall without choline, suggesting that their binding capacity is restricted to their specific interaction with this amino‐alcohol (Fig 4B). Notably, the isolated N‐CbpF was able to bind to the pneumococcal cell wall without choline, suggesting that N‐CbpF is able to specifically recognize the pneumococcal cell wall and that the N‐ and C‐terminal modules of CbpF bind to different components of the peptidoglycan.
The N‐terminal module of CbpF shows a markedly different three‐dimensional structure from the CBM. Electrostatic potential on the surface is also different in the two modules.
Docking analyses have shown that pneumococcal peptidoglycan can be accommodated perfectly at the N‐terminal cavity of CbpF (Fig 5), reinforcing the hypothesis that CbpF might act by blocking the access of LytC to its substrate. A fragment of the peptidoglycan ([GlcNAc‐MurNAc‐(l‐Ala‐γd‐Glu‐l‐Lys‐d‐Ala‐d‐Ala)]2) (2S5P)2 interacts with aromatic and hydrophobic residues of the cleft (Tyr 35, Tyr 48, Trp 75, Pro153 and Leu154) and with polar residues (Asp 38, Glu 50 and Asn 152; supplementary Table S2 online). Crystallization trials failed to provide direct information on peptidoglycan binding at the CbpF N‐terminal module. Native crystals fractured when put in contact with the synthetic cell wall fragment, and only in one case with a rapid soak could a diffraction data set be collected (supplementary Table S1 online). Despite the absence of electron density for peptidoglycan, the presence of the ligand caused a significant alteration of the CbpF structure at the CI domain and at the N‐terminal module (supplementary Fig S1 online). Long loops of the CI domain of the complexed structure are highly disordered and it was not possible to trace even the backbone between residues 168 and 178. The N‐terminal module in the presence of the peptidoglycan analogue rotated ∼10° around a molecular hinge between the N‐terminal module and the linker region, leading to large differences in the relative positions between the modules, and further supporting a direct interaction between the peptidoglycan and the N‐terminal module.
CbpF heads a protein family of unknown function
A study of the CbpF orthologues in other pneumococcal strains has showed the existence of proteins CbpC and CbpJ in the TIGR4 strain (supplementary Figs S3 and S4 online). Notably, a comparative analysis of the genes that encode these proteins shows that cbpF might have resulted from the recombination of cbpC and cbpJ. The coexistence of two highly similar genes in the TIGR4 strain indicates that the corresponding encoded CbpF‐like proteins have a crucial physiological function and that this duplication might favour the acquisition of more selective regulatory properties.
Further analyses of known pneumococcal genomes show the existence of several proteins that are likely to show the same architecture as CbpF. These proteins might constitute a new CbpF‐like subfamily, within the large CBP family of proteins, that have a typical CBM and an N‐terminal region formed by a series of non‐consensus repeats. Despite the minor differences in amino‐acid composition among all the members of the CbpF‐like subfamily, most of the residues involved in both the N‐terminal structural framework and in the intermodular interactions are preserved, indicating that these proteins might have similar regulatory functions to CbpF.
We have disclosed the structure and fold for CbpF, which is unique among the other functional domains of CBPs so far determined. This structure and the new bifunctional regulatory role that we have documented for CbpF open future avenues of research into the characterization of other pneumococcal members of this new subfamily of CBPs, the structure and function of which remain unknown.
Purification, crystallization and in vitro assays. CbpF was produced, purified and crystallized as described earlier (Molina et al, 2007). Detailed information about in vitro assays is described in the supplementary information online.
Structure determination. A gadolinium derivative and a data set at the peak of the Gd LIII absorption edge were obtained. SAD data were collected from frozen crystals at 100 K using a charged couple detector on beamline ID29 at the European Synchrotron Radiation Facility (ESRF). Data processing and structure determination is described in the supplementary information online.
Docking studies. The docking studies were carried out with the FlexiDock module of the SYBYL 6.9 suite of programs (Tripos Inc., St Louis, MO, USA). Detailed information about docking protocols is described in the supplementary information online.
Coordinates. The structure of CbpF–choline and CbpF–(2S5P)2 complexes have been deposited in the Protein Data Bank (accession codes 2v04 and 2vyu).
Supplementary information is available at EMBO reports online (http://www.emboreports.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
We thank E. García for a critical reading of the paper. This study was supported by grants from the Spanish Ministry of Science and Technology (BFU2005‐01645 and SAF2006‐00390) and by the COMBACT program of the Comunidad de Madrid (S‐BIO‐0260/2006) and CIBER de Enfermedades Respiratorias (CIBERES), which is an initiative of Spanish Instituto de Salud Carlos III. R.M. and A.G. hold a fellowship from the Spanish Ministry of Science and Technology. S.C. holds a Juan de la Cierva fellowship from the Spanish Ministry of Science and Technology. The study in the USA was supported by the National Institutes of Health.
- Copyright © 2009 European Molecular Biology Organization